Selective Serotonin Reuptake Inhibitor Use by Patients with Acute Coronary Syndromes

Patient Sample and Data Collection 

This study was approved by the Johns Hopkins Medicine Institutional Review Board. All patients who were admitted to The Johns Hopkins Hospital in Baltimore, Maryland from April 1, 2001 through April 30, 2004 with an acute coronary syndrome (ie, a diagnosis of unstable angina or of MI with or without ST-segment elevation) and who were treated with any glycoprotein IIb/IIIa receptor inhibitor during the hospitalization were identified with CView, a Johns Hopkins Health System hospital billing software application.16

Both electronic and handwritten medical records were reviewed. Recorded information included demographics, admission diagnosis, antidepressant use, glycoprotein IIb/IIIa inhibitor agent, medical comorbidities, risk factors for cardiovascular disease and bleeding, use of percutaneous transluminal coronary angioplasty, recurrent angina, MI during the hospitalization, asymptomatic cardiac enzyme elevation, bleeding, cardiac death, and total mortality. The data collection form also recorded the use of aspirin, clopidogrel, warfarin, heparin, nonsteroidal anti-inflammatory drugs, and direct thrombin inhibitors.

Outcome Measures 

Bleeding events and adverse cardiac events were monitored during the index hospitalization. Information was extracted from the handwritten and electronic medical records and recorded on an 8-page data collection form. Responses to these questions were then used to code whether a bleeding event or adverse cardiac event occurred. Major bleeding was defined as bleeding resulting in death; intracranial hemorrhage; retroperitoneal bleeding; or bleeding accompanied by a drop in the hemoglobin concentration of >3 g/dL.17, 18, 19 Minor bleeding was defined as overt bleeding not meeting criteria for major bleeding.19 This included any mention of blood loss in the medical record or the need for any intervention to treat a puncture site complication that resulted in bleeding. Blood loss resulting from coronary artery bypass graft surgery was not counted under major or minor bleeding. Gastrointestinal bleeding included any note of blood in the stool or vomitus (ie, melena, hematemesis, hematochezia). Any need for endoscopy during the hospitalization was documented, and then if this was needed, the endoscopy report was examined to determine whether bleeding was found at the time of the procedure.

Prior to chart review, adverse cardiac events were classified as “major” or “minor.” Major adverse cardiac events included death, MI during the hospitalization, and/or the need for urgent revascularization.20 Death was defined as mortality from any cause. An MI during the hospitalization was defined as an episode of typical ischemic chest pain accompanied by new elevation of the MB fraction of creatine phosphokinase with at least one value of 8 ng/mL or more or by a new elevation of the cardiac troponin I of greater than 3 times the upper limit of normal (in a patient with unstable angina as the admission diagnosis). For a patient with an admitting diagnosis of MI, an MI during the hospitalization was defined as typical ischemic chest pain and an increase in creatine phosphokinase-MB of at least 50% above the previous creatine phosphokinase-MB value with a characteristic rising and falling pattern of the creatine phosphokinase-MB fraction. In addition to major adverse cardiac events, minor adverse cardiac events were monitored, including recurrent myocardial ischemia (ischemic chest pain with new electrocardiographic changes but without elevation of cardiac biomarkers), new heart failure, and asymptomatic cardiac enzyme elevation (as defined above). Patients were classified hierarchically; if criteria for both major and minor adverse cardiac events were met, the patient was included only in the major adverse cardiac event group.

Statistical Analysis 

Statistical significance was defined as a probability value less than .05. Chi-square and Fisher’s exact tests were used to compare categorical variables and 2-tailed t tests to compare continuous demographic and clinical variables between patients who were administered an SSRI at any time during the hospitalization and patients who were not administered an SSRI.

Risk for bleeding and major adverse cardiac events was assessed by calculating odds ratios for patients who were administered an SSRI compared to patients not administered an SSRI. To adjust for differences in demographic and clinical characteristics of patients who were administered an SSRI in the hospital and those who were not administered an SSRI, we used a propensity score approach to generate a set of matched cases.21, 22 The propensity score is a measure of the likelihood that a patient received an SSRI based on relevant demographic and clinical variables. This approach reduces many covariates to a single variable, the propensity score. Since patients with similar propensity scores have similar characteristics, once matched, the outcome of interest can be compared for patients who received an SSRI versus those who did not receive an SSRI.

A stepwise logistic regression approach with P ≤.20 as the limit for variable entry was used to model the likelihood that patients received an SSRI based on demographic and clinical variables. A propensity score was generated for each patient, and each patient who received an SSRI during the hospitalization was matched with a patient with the closest possible propensity score who did not receive an SSRI. The maximum allowable difference between propensity scores for matching was set at 0.1. When 2 or more patients who did not receive an SSRI had an identical propensity score, the match patient was chosen randomly.23 All analyses were performed with SPSS version 13.0 (Chicago, Ill). Data are reported as means + standard deviation or as odds ratios with 95% confidence intervals (CI).

Sample Characteristics 

A total of 1254 patients were admitted with an acute coronary syndrome and received a glycoprotein IIb/IIIa inhibitor during the study period. Baseline demographic and clinical characteristics of the patients are shown in Table 1. The mean age of the sample was 63.1±12.0 years (range, 26 to 92), and 70.1% were male. Over half of the patients were admitted with unstable angina pectoris (55.2%), 26.6% were admitted with non-ST segment elevation MI, and 18.2% with ST segment elevation MI. Most patients had hypertension (76.4%) and a history of smoking (57.8%). Almost all patients underwent percutaneous transluminal coronary angioplasty (98.6%).

Table 1. Baseline Demographic and Clinical Characteristics

			Pre-Propensity Score Matching		Post-Propensity Score Matching
		SSRI (n=158)	No SSRI (n=1096)	P Value	No SSRI (n=158)	P Value
	Age (Years), Mean±SD	61.4±11.8	63.4±12.0	.06	61.2±12.1	.84
	Male Gender, n (%)	90(57.0)	789(72.0)	<.01	91(57.6)	.91
	Myocardial Infarction⁎, n(%)	49(31.0)	513(46.8)	<.01	55(34.8)	.47
	Hypertension†, n(%)	133(84.2)	825(75.3)	.01	133(84.2)	1.00
	Diabetes Mellitus†, n(%)	66(41.8)	345(31.5)	.01	65(41.1)	.91
	Hypercholesterolemia†, n(%)	128(81.0)	866(79.0)	.56	127(80.4)	.89
	Smoking†, n(%)	102(64.6)	623(56.8)	.07	109(69.0)	.40
	Chronic obstructive pulmonary disease†, n(%)	31(19.6)	124(11.3)	<.01	29(18.4)	.77
	Prior MI†, n(%)	68(43.0)	309(28.2)	<.01	64(40.5)	.65
	Congestive heart failure†, n(%)	35(22.2)	143(13.0)	<.01	36(22.8)	.89
	Left ventricular ejection fraction <35%, n(%)	26(17.7)	141(14.3)	.27	27(19.1)	.75
	Liver disease, n(%)	7(4.4)	18(1.6)	.02	5(3.2)	.56
	Renal failure†, n(%)	20(12.7)	81(7.4)	.02	22(13.9)	.74
	Gastrointestinal bleed†, n(%)	9(5.7)	41(3.7)	.24	5(3.2)	.27
	Anemia†, n(%)	20(12.7)	116(10.6)	.43	24(15.2)	.52
	Cancer†, n(%)	17(10.8)	123(11.2)	.86	10(6.3)	.16
	Bleeding diathesis†, n(%)	2(1.3)	4(0.4)	.17	0(0.0)	.50
	Depression†, n(%)	98(62.0)	80(7.3)	<.01	18(11.4)	<.01
	Non-SSRI antidepressant, n(%)	20(12.7)	80(7.3)	.02	19(12.0)	.86

SSRI=selective serotonin reuptake inhibitor; MI=myocardial infarction.

⁎ Admitting diagnosis. † Any history.

Approximately 1 of every 7 patients (14.2%) had a history of depression. A total of 158 patients (12.6%) were administered an SSRI during the hospitalization. Compared to patients who were not administered an SSRI, patients who received an SSRI during the hospitalization were significantly more likely to be female; to have been admitted with unstable angina pectoris; to have had a prior MI; and to have had a history of hypertension, diabetes mellitus, chronic obstructive pulmonary disease, congestive heart failure, renal failure, or liver disease. The majority of patients (62%) administered an SSRI during the hospitalization had a history of depression, whereas less than 10% of patients not administered an SSRI had a history of depression. Patients receiving an SSRI during the hospitalization were more likely to also be receiving another, non-SSRI, antidepressant than patients not receiving an SSRI (12.7 vs 7.3%, P =.02).

The propensity score model included 11 predictor variables (age, gender, admission diagnosis of MI, congestive heart failure, chronic obstructive pulmonary disease, prior MI, history of hypertension, history of smoking, history of renal failure, liver disease, and non-SSRI anti-depressant use) and 5 interactions (age×prior MI, age×chronic obstructive pulmonary disease, gender×chronic obstructive pulmonary disease, gender×history of hypertension, gender×history of smoking). The c-statistic for the propensity score model was 0.72, indicating adequate discrimination. Collinearity analysis showed that the variables included in the propensity score were not highly correlated; the highest bivariate correlation between any two variables was 0.22. All 158 patients who were administered an SSRI were successfully matched: 46 on 5 or more digits of the propensity score, 27 on 4 digits, 47 on 3 digits, 20 on 2 digits, and 18 on 1 digit. After propensity score matching, the only significant difference between the SSRI and non-SSRI patients was that patients administered an SSRI were more likely to have a history of depression (Table 1).

As shown in Table 2, almost all matched patients received aspirin (99.1%), clopidogrel (94.6%), and intravenous unfractionated heparin (98.3%) as standard therapy; 99.2% received either unfractionated or low molecular weight heparin. There were no significant differences between the patients administered an SSRI and those who were not, with respect to medications administered during the hospitalization that are known to affect bleeding or thrombosis risk. These medications included aspirin, clopidogrel, nonsteroidal anti-inflammatory drugs, warfarin, unfractionated heparin, low molecular weight heparin, glycoprotein IIb/IIIa inhibitors, or direct thrombin inhibitors. Similarly, there were no significant differences in the administration of these medications during the hospitalization after propensity score matching. The frequency of percutaneous transluminal coronary angioplasty was essentially the same among those administered an SSRI (156 of 158, 98.7%) and those who did not receive an SSRI (154 of 158, 97.5%, P=.68).

Table 2. Antiplatelet and Anticoagulant Medications during Hospitalization

	Medication		Pre-Propensity Score Matching		Post-Propensity Score Matching
		SSRI (n=158)	No SSRI (n=1096)	P Value	No SSRI (n=158)	P Value
	Abciximab, n (%)	8(5.1)	75(6.8)	.40	14(8.9)	.19
	Eptifibatide, n(%)	150(94.9)	1027(93.7)	.55	144(91.1)	.19
	Tirofiban, n(%)	1(0.6)	6(0.5)	.99	2(1.3)	.99
	Aspirin, n(%)	156(98.7)	1087(99.2)	.58	157(99.4)	.99
	Clopidogrel, n(%)	151(95.6)	1063(97.0)	.34	148(93.7)	.45
	Nonsteroidal anti-inflammatory drug, n(%)	4(2.5)	44(4.0)	.51	6(3.8)	.52
	Warfarin, n(%)	20(12.7)	106(9.7)	.24	15(9.5)	.37
	Unfractionated heparin, n(%)	153(96.8)	1080(98.5)	.12	157(99.4)	.21
	Low molecular weight heparin, n(%)	23(14.6)	149(13.6)	.74	17(10.8)	.31
	Lepirudin, n(%)	1(0.6)	6(0.5)	.99	0(0.0)	.99

SSRI=selective serotonin reuptake inhibitor.

Of the 158 patients who were administered an SSRI, sertraline was administered to 75 (47.5%), fluoxetine to 42 (26.6%), paroxetine to 26 (16.5%), escitalopram to 8 (5.1%), and citalopram to 7 (4.4%). In addition, 100 patients were administered a non-SSRI antidepressant; 46 received buproprion, 22 trazadone, 15 amitriptylline, 9 venlafaxine, and 15 either doxepine, imipramine, mirtazapine, nefazodone, or nortriptylline (7 patients were administered two non-SSRI antidepressants). Twenty patients who received an SSRI also received a non-SSRI antidepressant.

Adverse Cardiac Events and Bleeding 

Adverse cardiac event rates and bleeding outcomes for matched patients are summarized in Table 3. A total of 15 patients died during the index hospitalization (1.2%). Major and minor adverse cardiac events occurred in 86 patients (6.9%) and 163 patients (13.0%), respectively. A total of 98 patients experienced major bleeding (7.8%), 287 minor bleeding (22.9%), and 34 gastrointestinal bleeding (2.7%). Any bleeding occurred in 396 patients (31.6%). Patients who received an SSRI during the hospitalization were significantly more likely to experience any bleeding episode compared to those who did not receive an SSRI (37.3% vs 26.6%, OR 1.65, 95% CI, 1.02-2.66, P =.04). This was due principally to an increase in the risk of minor bleeding (27.2% vs 18.4%, OR 1.66, 95% CI, 0.98-2.84, P =.06). Of note, the bleeding rates observed in this study population are similar to those reported among patients with an acute coronary syndrome receiving conventional antiplatelet therapy and heparin.19, 24

Table 3. Adverse Cardiovascular and Bleeding Events During Hospitalization

	Event	No SSRI After Matching (n=158)	SSRI (n=158)	Odds Ratio (95% Confidence Interval)	P Value
	Major Adverse Cardiac Events	10(6.3%)	12(7.6%)	1.22(0.51-2.90)	.66
	Minor Adverse Cardiac Events	22(13.9%)	11(7.0%)	0.46(0.22-0.99)	.04
	Major Bleeding	13(8.2%)	11(7.0%)	0.84(0.36-1.92)	.67
	Minor Bleeding	29(18.4%)	43(27.2%)	1.66(0.98-2.84)	.06
	Gastrointestinal Bleeding	2(1.3%)	5(3.2%)	2.55(0.49-13.34)	.25
	Any Bleeding	42(26.6%)	59(37.3%)	1.65(1.02-2.66)	.04

SSRI=selective serotonin reuptake inhibitor. Patients who experienced both a major and a minor adverse cardiac event were only counted as having had a major adverse cardiac event (see text for definitions). A bleeding event was only counted once in the total if a person had more than one type (see text for definitions).

Patients who received an SSRI during the hospitalization were significantly less likely to experience a minor adverse cardiac event (ie, recurrent myocardial ischemia, heart failure or asymptomatic cardiac enzyme elevation) than individuals who did not receive an SSRI (7.0% vs 13.9%, OR 0.46, 95% CI, 0.22-0.99, P =.04). In the SSRI group, there were 5 patients with recurrent myocardial ischemia, 1 with heart failure, and 5 with asymptomatic cardiac enzyme elevation. Among those who did not receive an SSRI, there were 11 patients with recurrent myocardial ischemia (P =.12 vs SSRI group), 1 with heart failure (P =1.0), and 10 with asymptomatic cardiac enzyme elevation (P =.19 vs SSRI group). The likelihood of major adverse cardiac events did not differ between groups that did or did not receive an SSRI (7.6% vs 6.3%, P =.66).

All analyses were repeated for patients who were administered a non-SSRI antidepressant. The degree of balancing of demographic and clinical characteristics through propensity score matching was comparable to the SSRI analysis (not shown). There were 5 major adverse cardiac events among the 100 patients administered a non-SSRI antidepressant and 5 among matched controls. Compared to matched patients, however, patients who were administered a non-SSRI antidepressant during the hospitalization tended, albeit non-significantly, to experience more minor adverse cardiac events (14.0% vs 7.0%, OR 2.16, P =.11) compared to those patients who did not receive a non-SSRI antidepressant. Non-SSRI antidepressant use was not associated with bleeding (OR 1.04, P =.89) or, individually, with major (OR 1.09, P =.84), minor (OR 0.95, P =.87), or gastrointestinal (OR 0.59, P =.72) bleeding. Results for the SSRI and non-SSRI antidepressant analyses did not change substantively when the 20 patients who used both types of antidepressants were excluded from analyses.